⚠️ Your AI Can Already Do Things You Never Authorized

In this short, Microsoft Research leader Ece Kamar explains why today’s AI agents are far more unpredictable—and powerful—than most people realize. Unlike simple chatbots that only respond when prompted, modern AI agents can autonomously re-plan, reason through obstacles, and find new ways to achieve a goal, even if that means taking actions you never imagined: resetting passwords, hiring gig workers, or filing requests with government systems through tools and APIs. As their “action space” explodes, the risk boundaries we assume are safe no longer hold. Ece Kamar is the Managing Director of the AI Frontiers Lab at Microsoft Research, where she focuses on advancing responsible AI, agentic systems, and human–AI collaboration. In this conversation, she breaks down how rapidly evolving AI capabilities are redefining safety, control, and accountability—and why we urgently need new guardrails for this next wave of intelligent agents. 📺 Full episode: https://youtu.be/Cw5TOp53-5k 🎧 Podcast: https://www.buzzsprout.com/2234096/episodes/18153823 #startuptoexit #AI #AIagents #artificialintelligence #responsibleAI #AIsafety #futureofAI #autonomousagents #technology #machinelearning #generativeAI #ethicsinAI